Dynamic orders of decentralized H_inf controllers

The problem of decentralized control is addressed, i.e. the problem of designing a controller where each control input is allowed to use only some of the measurements. It is shown that, for such problems, there does not always exist a sequence of controllers of bounded order which obtains near-optimal control. Neither does there exist an infinite-dimentional optimal controller. Using the insight of the line of the proof of these results, a heuristic design algorithm is proposed for designing near-optimal controlers of increasing orders.